the impact of inward fdi on host countries: why such ... · host countries: why such different...

23

The Impact of Inward FDI on Host Countries: Why Such Different Answers?ROBERT E. LIPSEY and FREDRIK SJÖHOLM

2

A substantial body of literature has grown around the question of howinward foreign direct investment (FDI) affects host countries. On almostevery aspect of this question there is a wide range of empirical results inacademic literature with little sign of convergence. At the same time, policy-makers seem to have made their own judgments that inward FDI is val-uable to their countries. The United Nations Conference on Trade andDevelopment (UNCTAD) publishes annual data on “changes in nationalregulations of FDI” and reports that from 1991 through 2002, over 1,500changes making regulations more favorable and fewer than 100 makingregulations less favorable to FDI were made (UNCTAD 2003, 21, table 1.8).The same document reports that “the use of locational incentives to attractFDI has considerably expanded in frequency and value” (UNCTAD 2003,124). Given the amount of academic literature on the issue, why has it madeso little impression on policymaking? Are all these countries foolishly pur-suing an ephemeral fad? Are the questions asked in academic literatureirrelevant to policy? Are the relevant questions answerable? For that matter,what are the relevant questions?

Robert E. Lipsey is emeritus professor at the City University of New York and a research associate anddirector of the New York office of the National Bureau of Economic Research. Fredrik Sjöholm isassociate professor and researcher at the European Institute of Japanese Studies at the Stockholm Schoolof Economics.

2668-02_CH02.qxd 04/14/05 08:42 Page 23

24 DOES FDI PROMOTE DEVELOPMENT?

There are many possible effects of FDI inflow on a host country. Since itis generally taken for granted that investing firms possess some technol-ogy superior to that of host country firms, higher-quality goods and ser-vices could be produced at either lower prices or in greater volume thanpreviously available, resulting in higher consumer welfare. Another pos-sible effect would be that inward investment adds to the host country cap-ital stock, thereby raising output levels. Although this issue has beenexplored, especially in earlier literature determining whether inward invest-ment or aid supplements or displaces local investment, it is not specific todirect investment. Specific attention to direct investment has been devotedto the question of whether inward investments do involve superior tech-nology and, if they do, whether it “spills over” to domestically owned firmsrather than being retained entirely by the foreign-owned firms. A relatedset of questions is whether the foreign-owned firms pay higher wages fordomestic labor, whether those higher wages raise the average wage levelin the host country, and whether these higher wages spill over to domesti-cally owned firms. For both wages and productivity, the spillovers todomestically owned firms or establishments could be either positive ornegative. Wage spillovers could be negative if, for example, the foreign-owned firms hired the best workers, at their going—or higher—wages,leaving only lower-quality workers at the domestically owned firms.Productivity spillovers could be negative if foreign-owned firms took mar-ket shares from domestically owned firms, leaving the latter to produce atlower, less economical production levels.

Survey articles have found inconclusive evidence in the literature regard-ing the most important effects of inward FDI, especially with respect tospillovers. For example, on wage spillovers, Görg and Greenaway (2001)reported that panel data showed negative spillovers, while cross-sectionaldata reported positive spillovers. The same research paper found, withrespect to productivity spillovers from foreign-owned to domesticallyowned firms,“ only limited evidence in support of positive spillovers. . . .Most work fails to find positive spillovers, with some even reporting nega-tive spillovers . . .” (Görg and Greenaway 2001, 23). Görg and Strobl (2001)concluded that the crucial determinant of the findings in 21 studies waswhether cross-section or time-series data had been used, with the formertypically finding positive spillovers and the latter often negative ones.Lipsey stated that “the evidence for positive spillovers is not strong”(2003, 304) and concluded a review of the literature by saying that “theevidence on spillovers is mixed. No universal relationships are evident”(2004, 365). With respect to effects on host country economic growth,Carkovic and Levine (2002) found no significant effect of FDI inflows overthe entire 1960–95 period and only irregularly significant effects in five-yearintervals. None of the variables found in other studies consistently determinethe effect of FDI on growth, although some are significant in some combi-nation of conditioning variables. For instance, Lipsey found it “safe to

2668-02_CH02.qxd 04/14/05 08:42 4

conclude that there is no universal relationship between the ratio of inwardFDI flows to GDP and the rate of growth of a country” (2003, 297).

A crucial feature of these surveys is that the summarized studies donot individually find that wage or productivity spillovers do not exist.Mostly, they find evidence for either positive or negative spillovers. Inthis chapter, we try to understand why different investigators find con-tradictory results. Is it that the statistical techniques are different? Are thecountries they examine different? Are they asking different questionsunder the same labels of wages, productivity, or spillovers? We try toanswer these questions in two ways. One is to review the individual studiesthemselves to clarify the questions asked and the data used. The other isto survey studies on data for Indonesia, which cover a long period andare both detailed and accessible, in order to test the implications of dif-ferent definitions and methods. The studies we review in this chapterexamine the effects of FDI on firms and their workers. They are all produceroriented. However, future statistical studies could look at consumptioneffects. For example, has FDI growth in retailing reduced the price offood and other consumer goods? Has FDI growth in utilities reduced theprice of telephone service or home heating and lighting? These possibleeffects of FDI are almost totally absent from the literature but should bestudied.

Wage Spillovers

We begin with the studies of wage spillovers, which are not as numerousas those on productivity. There are several general issues that run throughalmost all the wage studies. One issue is that wage levels are calculated astotal wages or total compensation per worker, but the only measure ofskill is a division between production and nonproduction or blue-collarand white-collar workers. Within those categories, almost no studies candistinguish between differences in skill or education level or betweenemployees of foreign-owned and domestically owned plants from differ-ences in wages for identical workers. Similarly, they cannot distinguishbetween differential changes in skills between the two ownership groupsand differential changes in wages for identical and unchanging work-ers in plants owned by the two ownership groups. A second issue iswhether wage comparisons should take account of characteristics thatare correlated with foreign ownership but not intrinsically related to it.For example, foreign-owned firms or establishments are typically muchlarger on average than domestically owned ones, even in developedcountries. Especially in developing countries, foreign-owned firms orestablishments are more capital intensive and use more purchased mate-rials or components for their production than domestically owned firms.The question is whether these characteristics should be treated as con-

THE IMPACT OF FDI ON HOST COUNTRIES 25

2668-02_CH02.qxd 04/14/05 08:42 Page 25

trols—and their influence eliminated—or are they so bound up with for-eign ownership that they should not be controlled for? As Aitken,Harrison, and Lipsey (1996, 368) point out, a host country may not carewhether higher employee wages in foreign-owned plants result from thefact that they are foreign owned or from the fact that they are large anduse capital-intensive technology and/or import-intensive technology.Size, capital intensity, and import intensity may all be elements of theforeign-owned firm’s technology.

Empirical studies provide strong evidence of a wage premium in foreign-owned firms (Lipsey 2004). Foreign firms pay higher employee wagesin both developed and developing countries, after controlling for firm-specific characteristics. It is of course possible that high employee wages inforeign-owned firms are caused, or at least biased, by foreign takeovers ofhigh-wage domestic firms. In a recent study (Lipsey and Sjöholm 2002)—using a 25-year panel of Indonesian manufacturing establishment dataand lacking la-bor force education data, but including most of the typi-cal independent variables—we were able to lay this issue to rest, at leastfor this one country. Foreign-owned firms did tend to acquire domesticplants with higher than average blue-collar wages for their industries, butthe margins over the averages were far too small to account for the wagedifferential between domestically owned and foreign-owned plants. Thus,selectivity in take-overs could not account for the wage gap. Further evi-dence included the discovery that after a foreign takeover of a domesticallyowned plant, both blue-collar and white-collar wages rose strongly, inabsolute terms and relative to their industries. Takeovers of foreign-ownedplants by domestic firms had the opposite effect on wages, illustrating thatforeign takeovers, rather than takeovers in general, produced wageincreases. Econometric analyses using the whole panel of establishmentsfound large wage differences in favor of foreign firms at every level ofindustry and geographical detail, and the differentials remained large evenwhen plant characteristics, such as size and the use of purchased inputs,were introduced into the wage equations. The finding that employeewages were higher in foreign-owned plants and became higher whendomestically owned plants became foreign owned was not dependent onthe use of cross-section rather than panel data.

Although the literature on wage comparisons between foreign- anddomestically owned firms is large, relatively few studies examine the effectof FDI on wages in domestically owned firms. Görg and Greenway (2001)review six studies on wage spillovers and report that of those with conclu-sions, three panel studies found negative spillovers and two cross-sectionstudies found positive ones. They do not include the information that someof the cross-section estimates for Mexico and Venezuela also give negativecoefficients for spillovers, suggesting that the choice of cross-section or panelestimation may not be so crucial.


2668-02_CH02.qxd 04/14/05 08:42 Page 26

Other subsequent studies have reported more evidence that wage spill-overs occur. Figlio and Blonigen (2000) concluded that the effect of a largenew foreign investment in South Carolina on aggregate wage levels wasso large that it could not have been solely the result of the high employeewages in the foreign-owned plants but must have involved spillovers todomestically owned plants. Their study differed from most others becauseit concentrated on geographical effects, rather than the effects within theindustry of the investment. Indeed, in the only wage study we know of thatuses education as a measurement of the quality of the labor force (Lipseyand Sjöholm 2004b), we made a variety of calculations of spillovers ina cross section of Indonesian manufacturing establishments. Assumingnational labor markets within broad industry groups, we found significantwage spillovers to domestically owned plants. Assuming national labormarkets within narrower industry groups also revealed significant spill-overs, albeit smaller ones. In addition, assuming that an industry withinan individual province represented a labor market still revealed that spill-overs to domestically owned establishments occur. The combination ofhigher wages in foreign-owned plants and spillovers to domesticallyowned plants meant that higher overall wages were associated with foreign ownership. Further evidence that the distinction between cross-section and panel data studies is not the crucial determinant of results onwage spillovers can be found in Driffield and Girma (2002), which uses apanel of establishments in the UK electronics industry from the AnnualRespondents Database (ARD) from 1980 to 1992. Driffield and Girma foundintraindustry and intraregion wage spillovers from FDI on wages in gen-eral, and the effect was larger for skilled than for unskilled workers. Astudy by Girma, Greenaway, and Wakelin (2001), using firm, rather thanestablishment, panel data for almost 4,000 firms in the United Kingdomfrom 1991 to 1996, also found some evidence of wage spillovers. On aver-age, when spillovers were assumed to be identical across industries andfirms, Girma, Greenaway, and Wakelin found no significant evidence forthem. However, when the effects were permitted to vary across industries,wage spillovers were found and were higher in industries where the pro-ductivity gap between foreign and domestic firms was lower. One way thisstudy differs from our earlier research (Lipsey and Sjöholm 2002) is that itexcludes firms that changed ownership, thereby eliminating one way inwhich foreign ownership affects wages. The effects of shifts to foreign own-ership had been found in an earlier study to be positive in the UnitedKingdom, as they were in Indonesia.

The accumulation of studies since the earlier surveys seems to have putto rest the suspicion that the findings of wage spillovers were solely the resultof ignoring firm differences in cross-section studies, since the spilloversdid appear in panel studies. Something else must account for the negativespillovers or lack of spillovers found in some developing countries. Asidefrom Indonesia, the positive spillovers have been found most frequently in


2668-02_CH02.qxd 04/14/05 08:42 Page 27

developed countries. Even in the United Kingdom, large differences in pro-ductivity between foreign-owned and domestically owned firms reducedor eliminated spillovers. One possible cause for the negative results insome developing countries is that the gap between foreign-owned anddomestically owned firms is too large for one group to influence the other.Another possibility is that the labor markets in some developing countriesare too segmented for wages in one group to influence the other. If we com-pare Mexico and Venezuela, two countries reported to show negative wagespillovers from foreign firms, with Indonesia, the United Kingdom, and theUnited States, for which positive spillovers were found, labor market con-ditions do seem different. An “employment laws index” produced by theWorld Bank (2003), and based upon the work of Botero et al. (2003), had arange in which a high number indicated very restrictive labor laws on hir-ing, firing, and conditions of employment. On the basis of this index,Mexico and Venezuela were ranked among the most restrictive countries,with index numbers of 77 and 75, respectively. The United Kingdom wasrated at 28 and the United States at 22. Indonesia was in between at 57, notflexible by developed-country standards, but relatively flexible for a devel-oping country.

Another topic not always considered is how the relevant labor market isdefined. Most studies implicitly define a labor market as an industry—atwhatever level of detail industry is reported. Some define the market as anindustry within the narrowest geographical area at which industry data areavailable. That may be appropriate for some countries or industries, butthere may also be national labor markets within an industry, or local labormarkets that straddle many industries (national labor markets may alsostraddle many industries). These differences in defining the labor marketmay affect findings on spillovers. Therefore, consideration of the indus-try and geographic construction of FDI measures is needed, and the con-clusion might be different for wages from what it is for productivity. Forwages, the appropriate definition depends on the range of a labor marketwithin which wages tend to be equalized, or at least within which onefirm’s wages influence those in other firms. The answers might be differ-ent in different countries or industries and at different times. In a recent study(Lipsey and Sjöholm 2004b), we tested the effect of different definitions of alabor market by using different industry and geographic classifications toexamine the sensitivity of the results. They used FDI measures at two-,three-, and five-digit industry levels and at both the national and provincelevels to examine the effect of foreign presence on the wages in locally ownedIndonesian plants. The results for these various definitions of a labor marketare shown in table 2.1. The coefficients vary substantially, but they remainstatistically significant in all specifications.1 The largest coefficients are for


1. See Lipsey and Sjöholm (2004b) for the complete empirical specifications and results.

2668-02_CH02.qxd 04/14/05 08:42 Page 28

definitions of the relevant market as either national, at the two-digit industrylevel, or provincial, for all manufacturing industries combined. However,there is concern that these coefficients may represent the tendency of for-eign firms to move into either high-wage geographical locations or high-wage industries. Those possible biases are reduced by further geographicalbreakdown—by province—and by successively greater industry detail, cul-minating in breakdowns by five-digit industry and province. The coeffi-cients are greatly reduced in size, but remain strongly significant, showingmargins of a quarter for blue-collar and over a third for white-collar work-ers. The most detailed breakdown does not necessarily give the most accu-rate estimate of the effect of foreign firms’ presence, however. It may missthe effect of higher wages and increased employment in foreign-ownedestablishments in one industry or province on wages in other industriesand provinces—possibly a more important effect than any within the sameindustry and province. Even the more aggregate measures may understatethe wage effect because they are confined to manufacturing, ignoring anyimpacts on agriculture, services, and trade.

Productivity Spillovers

Many of the same issues that affect studies of wage spillovers occur in themuch larger body of literature studying productivity spillovers. In addi-tion, there are broader problems with the productivity measurements. The


Table 2.1 Wage spillovers in Indonesian manufacturingFDI variable Blue-collar wage White-collar wage

Two-digit national 1.07 1.04(21.83)*** (16.42)***

Three-digit national 0.28 0.34(6.20)*** (5.43)***

Five-digit national 0.16 0.35(7.48)*** (11.46)***

All sectors province 1.05 1.22(32.81)*** (28.27)***

Two-digit province 0.47 0.53(13.85)*** (12.26)***

Three-digit province 0.39 0.44(12.93)*** (12.12)***

Five-digit province 0.24 0.38(11.34)*** (13.12)***

*** = significance at the 1 percent level

Note: t-statistics are in parentheses.

Source: Lipsey and Sjöholm (2004b).

2668-02_CH02.qxd 04/14/05 08:42 Page 29

objective is often described as measuring the spillovers of technology, orknowledge, from foreign-owned to domestically owned firms. In order tosimplify measurement, technology is narrowly defined to measure laborproductivity, total factor productivity, or differences in production func-tions. All three are reflections of technology, but they may be both too broadand too narrow. The comparison of production functions, often cited as anideal method, assumes that there are no differences in technological knowl-edge involved in choices about factor combinations or plant size. Thus, theoperation of a large plant, as opposed to operation of a small plant, requiresno different technological mastery. The operation of a capital-intensiveplant requires no technological skill beyond that required for a labor-intensive plant. The use of intermediate inputs from abroad or from a par-ent company involves no technology beyond that of using locally availableinputs. These are all assumptions implicit in production function compar-isons, but if they are invalid, and locally owned plants do not have the tech-nological skill to operate at the scale and factor combinations of foreign-owned plants, true technological differences between foreign-owned anddomestically owned plants are hidden, disguised as differences in scale ofproduction or factor combination choices.

There is another respect in which the definitions of technology are toonarrow. For example, if foreign investors’ technological superiority con-sists of knowledge about consumers’ tastes in foreign markets, or aboutmarketing a product in local or foreign markets, this knowledge will notbe visible in productivity or production function comparisons. Rather, itwill be seen in comparisons of export performance, but those are a differ-ent set of literature not usually characterized as technology. A very dif-ferent type of study that takes a broad view of technology is exemplifiedby Dobson and Chia’s (1997) country studies for Asia, Rhee and Belot’s(1990) country- and industry-specific case studies of “the critical role oftransnational corporations (TNCs) in the transfer of technical, marketing,managerial know-how to developing countries,” and Moran’s (2001, 2002)many examples of technology transfer. All of these are basically case stud-ies of particular transfers of technology, but not confined to either intra-industry or interindustry transfers and not confined to specific measuresof technology. All of them find evidence for transfers of technology, but itis difficult to confront their evidence with the statistical studies describedlater in this chapter because the questions are so different. The case stud-ies ask whether there are examples in which technology was transferredfrom foreign-owned to domestically owned firms, and the answer is “yes.”In contrast, the statistical studies ask whether on average domestically ownedfirms gain in a particular measure of technology because foreign-ownedfirms operate in the same industry and the same country or the sameregion, and the answer is “not universally.” Both of these answers could beaccurate; neither one contradicts the other, because they are answers to dif-ferent questions.


2668-02_CH02.qxd 04/14/05 08:42 Page 30

Case studies offer great flexibility. The exact nature of the technologytransfer can differ from example to example, from industry to industry,and from country to country. The length of time for the transfer to occurand be measured need not be specified in advance and can vary widely.The transfer can be within an industry, to supplying industries, or to con-suming industries. This flexibility is an advantage of the case studymethod, but it comes at a cost: Firms that do not receive foreign technol-ogy are often omitted from case studies measuring the effect of transfers.Thus, the universe for measuring effects is not always delineated, and theuniverse from which the case studies are drawn is not always defined. Incontrast, statistical studies tend to be rigid in specifying the length of timeover which effects are measured (whether it is a year or a set number ofyears). They specify some particular definition of a technology transfer(perhaps ignoring other important dimensions), and whether differencesamong countries or industries are to be studied. Statistical studies assumethe relevance of some particular measure of FDI and some functionalform for its effects. The studies’ greatest advantage is that they tend toexamine effects on whole industries, including the unlucky or less com-petent losers, as well as the successes. With microdata, they can look atthe characteristics of firms changing ownership as well as those forcedout of an industry, those entering, and those remaining. A goal for casestudies might be to assemble a collection of unsuccessful ventures and tocompare them with successful ones, not only with respect to their owncharacteristics but also, even more importantly, with respect to countryand industry environments. Baranson’s (1967) book on Cummins’ expe-rience in India, for example, contains an analysis of the effects of import-substitution policies that can be compared with experiences under moreliberal trade regimes.

A general problem with productivity comparisons and spillover studies,compared to wage studies, is their greater need for data. Productivitystudies require output measures, usually sales or value added. Sales byforeign-owned firms, particularly exports, are frequently intracompanytransactions. The values may not be the same as market values, becausethere are many incentives to alter them to minimize tax liabilities, and theincentives may be very different for foreign-owned firms from any thatdomestically owned firms face. Any manipulation of sales values wouldaffect value added even more, and there are incentives to manipulate theprofit portion of value added in addition to those affecting sales values.Furthermore, since value added includes profits, it may fluctuate farmore over time than any physical measure of production. The use of pro-duction functions requires measures of capital input, which are oftenmissing from census data. If measures of capital input are present, theirmeaning is often questionable, especially in countries that have sufferedmajor inflations, because it is uncertain if and how historical values havebeen adjusted to current price levels. As with wage spillovers, the Görg


2668-02_CH02.qxd 04/14/05 08:42 Page 31

and Strobl (2001) and Görg and Greenaway (2001) surveys conclude thatthe negative results from panel data studies are more reliable than thosefor cross-sections, and that there is therefore little evidence of positivespillovers from FDI. However, a number of new studies of productivityspillovers based on panel data have appeared. As is true for wage spillovers,these find more evidence for positive spillovers than the earlier ones. Forexample, Haskel, Pereira, and Slaughter (2002) use a panel of UK manu-facturing plants between 1973 and 1992 and find a positive and robustspillover effect of inward FDI on productivity in local plants. Keller andYeaple (2003) also find positive and robust effects of inward FDI in theUnited States on productivity in US manufacturing plants between 1987and 1996. Girma, Greenaway, and Wakelin (2001), using the firm datadescribed above, find that there are spillovers and that they are greaterfor firms in sectors in which local firms are technologically comparableto the foreign firms. Labor productivity and total factor productivityspillovers are similar in size. As with wage spillovers, the accumulationof studies has eroded the basis for the hypothesis that the distinctionbetween cross-section and panel data studies explains the wide range offindings.

In their panel data study of Venezuela, Aitken and Harrison (1999)show what is probably the strongest evidence for negative productivityspillovers. A rise in the foreign share of ownership in a sector reduced theoutput of individual domestically owned establishments and reduced theirtotal factor productivity over one- to three-year periods. The first-year neg-ative effect was particularly severe for small domestically owned plants,suggesting that they were the least efficient and most vulnerable to com-petition from the increasing efficiency associated with rises in foreignownership. Since Venezuela had been a relatively closed economy to bothtrade and inward direct investment in manufacturing during this period,it might have accumulated a larger than average stock of small, competi-tively weak firms.

In another panel study of a relatively closed economy Kathuria (2000)used data for large firms in India from 1975–76 to 1988–89, before thecountry’s period of liberalization. Technical efficiency was measuredfrom a function with value added as the production measure and laborand capital as inputs, and was calculated as the distance between the firmand the most efficient firm in its industry. Spillovers were deemed tohave occurred if the dispersion of efficiency levels among domesticallyowned firms in the industries studied—in which foreign-owned firmswere the efficiency leaders—were reduced. The foreign source of thespillovers was measured in two ways: the extent of foreign participationin the industry, which was represented by the foreign-owned firms’ shareof sales, and the stock of cumulated purchases of foreign technology bylocal firms. Foreign participation had a negative effect on the dispersionof efficiency among domestically owned firms. This effect was inter-


2668-02_CH02.qxd 04/14/05 08:42 Page 32

preted by the author as indicating negative spillovers. Kathuria pointsout, however, that a negative spillover in these terms could result if boththe foreign firms and the domestically owned firms gained in efficiencybut the foreign-owned firms gained more—a result that would have beeninterpreted as a positive spillover in the Aitken and Harrison framework.The stock of foreign technological capital of the local firms was positivelyrelated to their gains in efficiency. When the sample was split between“scientific” and “nonscientific” industries, the spillover effects were con-fined to the “scientific” group but were offset by a positive coefficient forthe cross-product of foreign presence and the local firm’s research anddevelopment (R&D) effort. The interpretation was that R&D-intensivelocal firms might have gained, or lost less, from the foreign presence thanfirms that did less R&D.

Productivity Spillovers in Indonesia

One way of understanding the variety of results would be to apply the sametechniques to the identical types of data in different countries. Since we donot have access to data from many countries, we instead review studies ofIndonesia and test alternative methods on that country’s data. One advan-tage of using Indonesian data for experimentation is that Indonesia col-lects consistent microdata on its manufacturing industry and these data havebeen increasingly used by a number of authors for plant-level studies. Anumber of studies on Indonesia show that foreign plants have higher pro-ductivity than locally owned plants (Takii and Ramstetter 2003; Okamoto andSjöholm 2005) and that plants that change ownership from local to foreignownership increase their level of productivity (Anderson 2000). In addition,there are several plant-level studies on productivity spillovers from FDI inIndonesian manufacturing, which are summarized in table 2.2. The first threestudies on spillovers from FDI in Indonesia used cross-section analysis (seetable 2.2). For instance, Sjöholm (1999a) examined plants in 1980 and 1991and found both the level and growth of labor productivity to be higher forlocally owned plants in sectors with a high foreign share of output. Therewas no evidence of regional intraindustry spillovers from FDI, but someindications of regional interindustry spillovers.

Sjöholm (1999b) used the same data as his earlier study to examine pos-sible determinants of spillovers. The results suggested that spillovers werepositively affected by the technology gap between domestic and foreignplants and by the degree of competition within the sector. Blomström andSjöholm (1999) examined spillovers from FDI in 1991. Their study differedin design from the previous two mainly in the use of capital stocks ratherthan investment ratios to control for capital intensity. There were positivespillovers from FDI, and no differences in the spillovers from joint ventureswith minority or majority foreign ownership.


2668-02_CH02.qxd 04/14/05 08:42 Page 33

Blomström and Sjöholm (1999)

Sjöholm (1999a)

Sjöholm (1999b)

Takii (2001)

Todo and Miyamoto(2002)

Blalock and Gertler (2002)

Blalock and Gertler (2003)

*** = significance at the 1 percent level+ = positive? = not statistically significant

34

Table 2.2 Studies on productivity spillovers from FDI in Indonesian manufacturing

Measure of t-statisticsDependent foreign Independent for foreign

Author(s) Year(s) variable presence variables share

1991

1980, 1991

1980, 1991

1990–95

1995–97

1988–96

1988–96

Value addedperemployee

Growth invalue added;value addedperemployee

Growth invalue added;value addedperemployee

Value added

Value addedperemployee

Output

Output

Output (five-digitlevel)



Employment(three-digitlevel)

Absoluteamount ofFDI output(two-digitlevel)

Output (four-digitlevel;region-industry)

Output (four-digitlevel;region-industry)

CapitalWhite/blueCapital

utilizationScale

Independentdummies

EmploymentInvestmentIndustry and

regional characteristics

EmploymentInvestmentScale

(translog)EmploymentCapitalPlant-specific

effect

CapitalCapacity

utilizationPlant-specific

effect

(translog)EmploymentCapitalRaw materialsPlant-specific

effect

(translog)EmploymentCapitalRaw materialsEnergyDownstream FDIPlant-specific

effect

+ ***

+ ***

+ ***

+ ***

+ ***

+ ***

?

2668-02_CH02.qxd 04/14/05 08:42 Page 34

Takii (2001) was the first study on spillovers in Indonesia that used paneldata, which allowed him to control for plant-specific effects. He examinedspillovers during 1990–95 using a translog production function and foundpositive effects on value added in local firms from the share of foreignemployment in the same three-digit International Standard IndustrialClassification (ISIC) industry. Moreover, the results suggested that spill-overs were relatively large in sectors with relatively new foreign plants andwith low gaps in labor productivity between foreign and domestic plants.Takii also found that R&D positively affected spillover in locally ownedplants.

The study by Todo and Miyamoto (2002) differs from most of the othersby defining the FDI variable as the absolute amount of FDI in a sector.They argued that this measure is more strongly related to the foreignknowledge stock and therefore preferred over the foreign share of a sector.The result showed a positive effect of FDI on local firms’ labor productivityafter controlling for R&D and training of the workforce.

Blalock and Gertler (2002) also used a translog production function toexamine spillovers between 1988 and 1996. Local firms in sectors withinregions with a high foreign share of output had high levels of productivity.Moreover, they found a positive effect on spillovers from the technology gapbetween domestic and foreign plants; spillovers were also positively affectedby local firms’ R&D and by high levels of education of workers in local firms.

In a second study, Blalock and Gertler (2003), using the same data and avery similar translog production function, found no evidence of positiveintraindustry spillovers from FDI. A second measure of FDI in this studyis the main difference between the two: Blalock and Gertler (2003) mea-sured FDI in upstream markets to capture spillovers from FDI to local sup-pliers. They found that downstream FDI was highly significant in theeconometric estimations. This variable was constructed by using an input-output table at a sector level, which also includes purchases from its ownsector. Therefore, one possibility is that the variable on downstream FDIalso captured the effect of horizontal spillovers.

To summarize the results from these seven production spillover studies onIndonesian manufacturing, all cross-section studies and three out of fourpanel data studies found statistically significant intraindustry spillovers. Theone study that failed to find intraindustry spillovers found interindustryspillovers from FDI instead. Judging by these studies of Indonesia, we con-clude that the design of econometric studies does not cause the differentresults found in the literature. Therefore, differences between countries orfirms may explain the extent of spillovers. The studies on Indonesia mightshed some further light on what these differences could be. Previous litera-ture suggests that competition, the technology gap, and local firms’ absorp-tive capacity will affect the extent of spillovers. Starting with competition, thestudies by Sjöholm (1999b) and by Blalock and Gertler (2003) show thatspillovers are highest in sectors with high competition. The former study sug-


2668-02_CH02.qxd 04/14/05 08:42 Page 35

gests that it is domestic competition, as captured by a Herfindahl index, ratherthan the degree of protection from imports that affects spillovers. The secondstudy suggests that competition will benefit upstream local suppliers.

The effects of technology gaps on the extent of spillovers is unclear.Takii (2001) found a negative effect on spillovers from the technology gapbetween local and foreign-owned plants, which has also been found in other countries (Kokko 1994, 1996). Sjöholm (1999b) and Blalock andGertler (2002) find a positive relation between the technology gap and thedegree of spillovers. One explanation for the different results could bethat the measure of technology gap differs between studies. Takii mea-sured the technology gap as the difference in labor productivity betweendomestically owned and foreign-owned plants.2 Sjöholm used the differ-ence in labor productivity between domestically owned and foreign-owned plants after controlling for the scale of operation and the investmentper worker ratio.3 Finally, Blalock and Gertler used the plant’s fixed effectin comparison to the mean fixed effect in the same industry. Anotherreason why these, and other studies, produce such varying results couldbe that the relationship is nonlinear. Some technology gap is presumablyrequired for any useful technology spillover to occur. However, it is alsoplausible that if the gap is too large, the technology in foreign plants willbe of little practical use in locally owned plants pursuing very differenttypes of operations.

Differences in spillovers between countries may also be caused by differences in sectors’ and plants’ absorptive capacity. The studies on Indo-nesia confirm that such capacity might be important if a firm is to benefitfrom spillovers. Takii (2001) and Todo and Miyamoto (2002) as well as Blalockand Gertler (2002) found that a firm’s own R&D positively affected its abilityto benefit from spillovers. The last study also found that plants with morehighly educated employees benefit more from the presence of foreign multi-national corporations (MNCs). A related question is whether the type of activ-ities pursued by the foreign subsidiaries affects spillovers to domesticallyowned firms. This issue has been rather neglected in the spillover literaturebut Todo and Miyamoto (2002) find a positive effect on spillovers fromR&D and human resource development in the foreign subsidiaries.

As evident from the earlier discussion, considerable attention has beendevoted to differences between econometric methodologies as one possi-ble explanation for the different effects of spillovers among countries. Arelated, but so far rather neglected, issue is how one should construct mea-sures of FDI. Most studies use the foreign share of a sector’s economic


2. Takii also used the difference in capital labor ratios and the difference in size as alternativemeasures of technology gaps. These measures gave inconclusive results.

3. The difference in investment ratios was used as an alternative measure but provided noclear results.

2668-02_CH02.qxd 04/14/05 08:42 Page 36

activity as a measure of FDI.4 One problem with this measure is that theforeign share of a sector might be endogenously determined if produc-tivity spillovers expand activity in local firms. Moreover, this measureassumes that increases of foreign and aggregate activity in the same pro-portion have no effect on local firms. Castellani and Zanfei (2002) arguethat this assumption might produce a downward bias on the estimate ofspillovers from FDI. Finally, it is not clear why we would assume the effectfrom FDI to be linear in the foreign share of an industry’s economic activ-ity: spillovers are not obviously maximized at a 100 percent foreign own-ership share (Lipsey 2004).

Although the foreign share is widely used as an FDI measure, produc-tivity spillover studies still differ in how this share is constructed. Somemeasure it as the foreign share of employment, while others measure it asa share of value added or output. Moreover, the foreign share is calculatedat different sector levels, ranging from two-digit to five-digit levels of ISIC.Finally, some studies use the foreign industry share at a national level,while others use it at a regional level.

The more narrow the definition of an industry, the more restrictive is ourassumption of how widely applicable knowledge from FDI can be for localfirms, and our assumption of which domestically owned plants face in-creased competition from FDI. If we construct the FDI measure on a two-digit level of ISIC, it implies that productivity spillovers might be presentbetween industries at a three- and five-digit level of ISIC but not from onetwo-digit industry to another. If we construct our measure of FDI at a five-digit level of ISIC, it implies that productivity spillovers can only be capturedif they occur within these industries but not if they cross from one five-digitindustry to another. It is unclear what a properly defined industry is for ananalysis of productivity spillovers. It seems that most studies favor a dis-aggregated definition of FDI, possibly to increase the variance in the FDIvariable. However, this might come at a cost if we miss out on spilloversacross narrowly defined industries. Some technologies, such as computeruse in tracking sales and inventories, may be very general and easily trans-mitted across industries, while others may be specific to particular produc-tion processes. Clearly, the industry definition will also have implications forwhat we attribute to interindustry versus intraindustry spillovers.

The choice to construct the FDI measure at a national level or at a regionallevel might also be important. Choosing the most appropriate level to usedepends on whether the spillover has a spatial dimension—for example, ifit primarily benefits plants within the same region. The Jaffe, Trajtenberg,and Henderson (1993) study is often referred to when a regional measure


4. There are exceptions, see, for example, the previously discussed study by Todo andMiyamoto (2002). See also Barrell and Pain (1997), who use aggregate FDI in a constant elas-ticity of substitution (CES) production function and find positive effects from FDI on techni-cal progress in EU countries.

2668-02_CH02.qxd 04/14/05 08:42 Page 37

of FDI is used (Sjöholm 1999a, Blalock and Gertler 2002, Lipsey andSjöholm 2004b). Their study shows that university R&D primarily bene-fits other inventors within the same geographic area. Hence, their studyrelates to innovation, and it is possible, but not certain, that the sameresult also exists for spillovers. Whether or not spillovers are geographi-cally concentrated depends on, for instance, whether imitation, competition,or supply of linkage industries are enhanced by geographic proximity to theforeign firms.

If we believe that technology spillovers are geographically concentrated,the next question will be: What is an appropriate geographic aggregationlevel? Studies on Indonesia have used both districts (Sjöholm 1999a) andprovinces (Sjöholm 1999a; Blalock and Gertler 2002, 2003). One method-ological problem is that spillovers are not likely to follow administrativeunits even if they are localized. For instance, the largest share of Indonesianmanufacturing is located in the province of Western Java. This is largelybecause the industry sector has grown out of its original base in Jakarta.Jakarta and the West Java cities of Bogor, Tanggerang, and Berakasi con-stitute one industrial cluster, the Jabotabek area (Henderson, Kuncoro, andNasution 1996). If technology spillovers from FDI exist, and even if suchspillovers are only effective with geographic proximity, a foreign firm inJakarta is likely to have positive effects on local firms within the wholeJabotabek area. However, Jabotabek spreads out over two provinces andabout ten districts, which indicates the problem of using administrativegeographic units in constructing measures of regional FDI.

Spatial concentration of FDI may be another obstacle to analyzing regionalFDI measures. However, such concentration is common in most countries,including Indonesia. For instance, about 80 percent of all FDI in Indonesianmanufacturing is located in 3 out of 27 provinces (East Java, West Java, andJakarta), which is a higher concentration than for manufacturing in general(Sjöholm 2002, Sjöberg and Sjöholm 2004). If, for instance, we construct ourFDI measure at a province level and at a five-digit level of ISIC—includingabout 300 industries—less than 25 percent of the region-industry combi-nations will have FDI. Thus, it may be desirable to take account of the selec-tion of locations in analyzing the effects of FDI.

An experiment with different industry and geographical definitions ofthe relevant scope for productivity spillovers is described in table 2.3.Spillovers are estimated at the national level and the province level for allsectors combined and at two-, three-, and five-digit industrial breakdowns.More specifically, we used Indonesian plant level data for 1996 to estimatethe following expression:

Laborprodij = constant + FDI + Educationij + Capitalij + Sizeij + Publicij

where Laborprod is value added per employee, Capital is energy consumptionper employee, Size is the total number of workers, Public is a dummy variable


2668-02_CH02.qxd 04/14/05 08:42 Page 38

for public ownership, and Education is the share of employees with primary,junior, senior, and university education for both blue- and white-collarworkers. For the sake of clarity, we show only the coefficients of the differentFDI variables in table 2.3.

The main impression from the results in table 2.3 is that geographicalinfluences are minor; the spillover coefficients at the national level arealmost identical to those at the province level at each level of industrydetail. The industry level does make a difference. The coefficient is high-est at the all-sector level, indicating a greater influence of foreign presenceon domestic establishment productivity for manufacturing as a wholethan within two-, three-, or five-digit industries. The coefficient is higherat the three-digit level than at the two-digit level, as one would expect ifspillovers tended to be largest within a narrow industry. However, theeffect becomes smaller when we move to the five-digit industries. Thebehavior of productivity spillovers contrasts with that of wage spillovers(table 2.1), where going from the national to the province level raised thespillover coefficient at the three-digit and five-digit industry levels. Thedifference between the wage and productivity spillovers is mostly, althoughnot entirely, consistent with the idea that wage spillovers come throughcompetition for labor in geographically narrow labor markets, while pro-ductivity spillovers result from competition in countrywide productmarkets.


Table 2.3 Productivity spillovers in Indonesian manufacturing (dependent variable:value added per employee)

FDI variable Coefficient of FDI

Two-digit/national 0.28(3.94)***

Three-digit/national 0.44(5.55)***

Five-digit/national 0.19(5.25)***

All sectors/province 0.94(19.05)***

Two-digit/province 0.27(5.44)***

Three-digit/province 0.44(9.79)***

Five-digit/province 0.23(6.44)***

*** = significance at the 1 percent level

Note: t-statistics are in parentheses.

Source: Authors’ calculations.

2668-02_CH02.qxd 04/14/05 08:42 Page 39

Conclusions

Why do studies of spillovers reach such diverse conclusions? With respectto wage spillovers, the use of cross-section or panel data does not seem todetermine the result. As far as we can judge from Indonesia, the tendencyof foreign-owned firms to gravitate to high-wage industries, while it exists,does not explain the apparent spillovers and neither does any tendency offoreign firms to take over high-wage local firms within industries. Asidefrom Indonesia, most of the evidence for wage spillovers comes fromdeveloped countries, particularly the United States and the United King-dom. One hint that differences in labor market institutions might be impor-tant for the degree of wage spillovers is that two countries found to havenegative spillovers were countries with very restrictive labor laws, whilethe United States and the United Kingdom were among the least restric-tive. With respect to productivity spillovers, an accumulation of panel datastudies has erased the previous unanimity of panel data results in show-ing negative or no spillovers. As with wages, firm-specific characteristicsdo not explain all the higher productivity found for domestic firms inindustries where foreign-owned firms were important. The econometricmethod does not seem to be the crucial determinant of the result.

An explanation that seems plausible at this point is that countries andfirms within countries might differ in their ability to benefit from the pres-ence of foreign-owned firms and their superior technology. There might becountries or industries in which the domestically owned sector is too smallor unable to learn from foreign-owned firms. In those cases, the domesticsector may be crowded out by competition from the more efficient foreign-owned firms. The state of the domestically owned sector might depend notonly on the stage of development of the economy, but also on the type oftrade regime. A heavily protected domestically owned sector might beinefficient and lacking in entrepreneurship. It makes sense that the arrivalof foreign firms with technology greatly superior to that of domesticallyowned firms should inflict damage on at least some domestic firms. Theleast efficient, perhaps often the smallest, might become unprofitable orbe forced out of the industry. One might view that outcome as favorablefor the host country as a whole if the average productivity of foreign-owned and domestically owned firms together increased. Few studies takeaccount of both the exit and the entrance of new firms, both of which areimportant for assessing the overall impact of inward FDI.

If country and industry differences are important to the impact ofinward FDI on host countries, the main lesson might be that the search foruniversal relationships is futile. In that case, the question shifts from howinward FDI affects every host country and industry to which types ofindustries and host countries are affected, and what the impact is on each.It is in identifying the characteristics of firms, industries, and countriesthat promote the transfer of technology that case studies can be most


2668-02_CH02.qxd 04/14/05 08:42 Page 40

valuable. Their flexibility with respect to assumptions regarding timingand types of technology transfer suggests what statistical studies shouldlook for and how the variables should be defined, especially if they encom-pass a wide range of both successful and unsuccessful ventures.

Why has academic skepticism about the impact of FDI not influencedpolicy more strongly? One reason is probably the diversity of findings.Another is the narrow scope of technology in the statistical tests. It relieson the assumption that the scale of operations and the import of compo-nents from abroad, and particularly from other related firms, do not con-stitute part of affiliate technology, but are simple inputs, accessible to localas well as foreign firms. Policymakers may have found these assumptionsimplausible.

References

Aitken, Brian J., and Ann E. Harrison. 1999. Do Domestic Firms Benefit from Direct ForeignInvestment? Evidence from Venezuela. American Economic Review 89, no. 3 (June): 605–18.

Aitken, Brian J., Ann E. Harrison, and Robert E. Lipsey. 1996. Wages and Foreign Ownership:A Comparative Study of Mexico, Venezuela, and the United States. Journal of InternationalEconomics 40, no. 3-4 (May): 345–71.

Anderson, Gary W. Jr. 2000. Multinational Corporations and Tacit Knowledge: Determinationof Entry Mode and Impact of Entry. Paper presented at the 7th Convention of the EastAsian Economic Association, Singapore, November 17–18.

Baranson, Jack. 1967. Manufacturing Problems in India: The Cummins Diesel Experience. Syracuse,NY: Syracuse University Press.

Barrell, R., and N. Pain. 1997. Foreign Direct Investment, Technological Change, and EconomicGrowth within Europe. Economic Journal 107, no. 445 (November): 1770–86.

Blalock, Garrick, and Paul J. Gertler. 2002. Firm Capabilities and Technology Adoption: Evidencefrom Foreign Direct Investment in Indonesia. Working Paper, Department of AppliedEconomics and Management, Cornell University. Ithaca, NY: Cornell University.www.aem.cornell.edu/faculty_sites/gb78/ (accessed January 3, 2004).

Blalock, Garrick, and Paul J. Gertler. 2003. Technology from Foreign Direct Investment and WelfareGains through the Supply Chain. Working Paper, Department of Applied Economics andManagement, Cornell University. Ithaca, NY: Cornell University. www.aem.cornell.edu/faculty_sites/gb78/ (accessed January 3, 2004).

Blomström, Magnus, and Fredrik Sjöholm. 1999. Technology Transfer and Spillovers: DoesLocal Participation with Multinationals Matter? European Economic Review 43, no. 4–6(April): 915–23.

Botero, Juan, Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes, and AndreiShleifer. 2003. The Regulation of Labor. NBER Working Paper 9756. Cambridge, MA:National Bureau of Economic Research.

Carkovic, Maria, and Ross Levine. 2002. Does Foreign Direct Investment Accelerate EconomicGrowth? Department of Finance Working Paper (June). Minneapolis: University ofMinnesota.

Castellani, Davide, and Antonello Zanfei. 2002. Multinational Companies and ProductivitySpillovers: Is There a Specification Error? ISI Università di Urbino. Photocopy.

Dobson, Wendy, and Chia Siow Yue, eds. 1997. Multinationals and East Asian Integration.Ottawa, Canada: International Development Research Centre.

Driffield, N., and Sourafel Girma. 2002. Regional Foreign Direct Investment and Wage Spillovers:Plant-Level Evidence from the Electronics Industry. Research Paper 2002/04. Nottingham,UK: Leverhulme Centre for Research on Globalisation and Economic Policy.


2668-02_CH02.qxd 04/14/05 08:42 Page 41

Figlio, David N., and Bruce A. Blonigen. 2000. The Effects of Foreign Direct Investment onLocal Communities. Journal of Urban Economics 48, no. 2 (September): 338–63.

Girma, Sourafel, David Greenaway, and Katherine Wakelin. 2001. Who Benefits from ForeignDirect Investment in the UK? Scottish Journal of Political Economy 48, no. 2 (May):119–33.

Görg, Holger, and David Greenaway. 2001. Foreign Direct Investment and Intra-industrySpillovers: A Review of the Literature. Globalisation and Labor Markets ProgrammeResearch Paper No. 2001/37. Nottingham, UK: Leverhulme Centre for Research onGlobalisation and Economic Policy.

Görg, Holger, and Eric Strobl. 2001. Multinational Companies and Productivity Spillovers:A Meta-analysis. Economic Journal 111, no. 475 (November): F723–39.

Haskel, Jonathan E., Sonia C. Pereira, and Matthew J. Slaughter. 2002. Does Inward ForeignDirect Investment Boost the Productivity of Domestic Firms? NBER Working Paper 8724.Cambridge, MA: National Bureau of Economic Research.

Henderson, Vernon J., Ari Kuncoro, and Damhuri Nasution. 1996. The Dynamics of JabotabekDevelopment. Bulletin of Indonesian Economic Studies 32, no. 1 (April): 71–95.

Jaffe, Adam B., Manuel Trajtenberg, and Rebecca Henderson. 1993. Geographic Localizationof Knowledge Spillovers as Evidenced by Patent Citations. Quarterly Journal of Economics108, no. 3 (August): 577–98.

Kathuria, Vinish. 2000. Productivity Spillovers from Technology Transfer to Indian Manufac-turing Firms. Journal of International Development 12, no. 3: 343–69.

Keller, Wolfgang, and Stephen R. Yeaple. 2003. Multinational Enterprises, International Trade,and Productivity Growth: Firm Level Evidence from the United States. NBER Working Paper9504. Cambridge, MA: National Bureau of Economic Research

Kokko, Ari. 1994. Technology, Market Characteristics, and Spillovers. Journal of DevelopmentEconomics 43, no. 2 (April): 279–93.

Kokko, Ari. 1996. Productivity Spillovers from Competition Between Local Firms and ForeignAffiliates. Journal of International Development 8, no. 4 (April): 517–30.

Lipsey, Robert E. 2003. Foreign Direct Investment, Growth, and Competitiveness in DevelopingCountries. In The Global Competitiveness Report, 2002–2003, ed., Peter K. Cornelius. New York:Oxford University Press.

Lipsey, Robert E. 2004. Home- and Host-Country Effects of Foreign Direct Investment. InChallenges to Globalization, ed., Robert E. Baldwin and L. Alan Winters. Chicago: Universityof Chicago Press.

Lipsey, Robert E., and Fredrik Sjöholm. 2002. Foreign Firms and Indonesian ManufacturingWages: An Analysis with Panel Data. NBER Working Paper 9417 (December). Cambridge,MA: National Bureau of Economic Research.

Lipsey, Robert E., and Fredrik Sjöholm. 2004a. Foreign Direct Investment, Education, andWages in Indonesian Manufacturing. Journal of Development Economics 73, no. 1 (February):415–22.

Lipsey, Robert E., and Fredrik Sjöholm. 2004b. FDI and Wage Spillovers in IndonesianManufacturing. Review of World Economics 140, no. 2: 287–310.

Moran, Theodore H. 2001. Parental Supervision: The New Paradigm for Foreign Direct Investment.Washington: Institute for International Economics.

Moran, Theodore H. 2002. Beyond Sweatshops: Foreign Direct Investment and Globalization inDeveloping Countries. Washington: Brookings Institution Press.

Okamoto, Yumiko, and Fredrik Sjöholm. 2005. FDI and the Dynamics of Productivity inIndonesian Manufacturing. Journal of Development Studies 41, no. 1 (January): 160–82.

Rhee, Yung Whee, and Therese Belot. 1990. Export Catalysts in Low-Income Countries. WordBank Discussion Paper 72. Washington: World Bank.

Sjöberg, Örjan, and Fredrik Sjöholm. 2004. Trade Liberalization and the Geography ofProduction: Agglomeration, Concentration and Dispersal in Indonesia’s ManufacturingIndustry. Economic Geography 80, no. 3 (July): 287–310.

Sjöholm, Fredrik. 1999a. Productivity Growth in Indonesia: The Role of Regional Characteristicsand Direct Foreign Investment. Economic Development and Cultural Change 47, no. 3 (April):559–84.


2668-02_CH02.qxd 04/14/05 08:42 Page 42

Sjöholm, Fredrik. 1999b. Technology Gap, Competition and Spillovers from Direct ForeignInvestment: Evidence from Establishment Data. Journal of Development Studies 36, no. 1(October): 53–73.

Sjöholm, Fredrik. 2002. The Challenge of FDI and Regional Development in Indonesia. Journalof Contemporary Asia 32, no. 3: 381–93.

Takii, Sadayukii. 2001. Productivity Spillovers and Characteristics of Foreign Multinational Plantsin Indonesian Manufacturing 1990–95. ICSEAD Working Paper 2001-14. Kitakyushu,Japan: ICSEAD.

Takii, Sadayukii, and Eric D. Ramstetter. 2003. Employment, Production, Labor Productivity, andForeign Multinationals in Indonesian Manufacturing, 1975–2000. ICSEAD Working Paper2003-25. Kitakyushu, Japan: ICSEAD.

Todo, Yasuyuki, and Koji Miyamoto. 2002. Knowledge Diffusion from Multinational Enterprises:The Role of Domestic and Foreign Knowledge-Enhancing Activities. OECD Technical Paper196. Paris: OECD Development Centre.

UNCTAD. 2003. World Investment Report 2003. UNCTAD: New York and Geneva.World Bank. 2003. Doing Business. Washington: World Bank.


2668-02_CH02.qxd 04/14/05 08:42 Page 43

2668-02_CH02.qxd 04/14/05 08:42 Page 44

the impact of inward fdi on host countries: why such ... · host countries: why such different...

Documents